Professional Documents
Culture Documents
* eivind.aadland@hisf.no
Abstract
Background
In adults, a minimum of 35 days of accelerometer monitoring is usually considered
appropriate to obtain reliable estimates of physical activity (PA). However, a longer period
of measurement might be needed to obtain reliable estimates of sedentary behavior (SED).
The aim of this study was to determine the reliability of objectively assessed SED and PA in
adults.
Methods
OPEN ACCESS Eighty-seven adult subjects (28 men; mean (standard deviation) age 31.3 (12.2) years;
Citation: Aadland E, Ylvisker E (2015) Reliability of body mass index 23.7 (3.1) kg/m2) wore the GT3X+ accelerometer for 21 subsequent days,
Objectively Measured Sedentary Time and Physical for which the reliability of different wear time criteria (8 to 12 h/day and 3 to 5 d/week) was
Activity in Adults. PLoS ONE 10(7): e0133296.
doi:10.1371/journal.pone.0133296
explored. Variance partitioning along with the Spearman-Brown prophecy formula was
used as the basis for determining intraclass-correlation coefficients (ICC) and the number of
Editor: Guillermo Lpez Lluch, Universidad Pablo de
Olavide, Centro Andaluz de Biologa del Desarrollo-
monitoring days needed (N) to achieve an ICC = 0.80. Week-by-week reliability was
CSIC, SPAIN reported using ICC, Bland-Altman plots and absolute measures of agreement.
Received: March 20, 2015
week-by-week variability was found for all variables. Researchers need to be aware of sub-
stantial intra-individual variability in accelerometer-measurements.
Introduction
Accelerometry has gained widespread popularity as a valid tool for the assessment of free-living
physical activity (PA) and sedentary behavior (SED) in diverse study settings. Yet, given the
inherent variation in behavior over time, an important aspect of accelerometer measurements
is how many days of measurement to be considered to obtain reliable results. Importantly,
whether PA is measured as a predictor (x-) variable or an outcome (y-) variable, noise will
hamper researchers ability to make valid causal inferences in PA epidemiology [1] and possi-
bly misinform the society regarding targets for public health management.
In adults, 35 days of monitoring are normally considered appropriate, which is in accor-
dance with recommendations given [2]. However, estimates of how many days of monitoring
that should be included to obtain a reliable result vary considerably between studies [37, 2, 8],
and might also vary between outcome variables of interest [6, 8]. According to Matthews et al
[6], inclusion of more days may be needed to arrive at reliable estimates (intraclass correlation
coefficient (ICC) 0.80) for physical inactivity (<500 cpm from the Actigraph 7164) (7
days), compared to PA ( 34 days). A comparable finding has been shown in older adults,
where 23 days was needed for PA, whereas 5 days of monitoring was needed for SED (<50
cpm from the Actigraph 7164). The possible impaired reliability for SED compared to other
variables may be of critical importance, given the increased interest in SED in the primary and
secondary prevention of a range of chronic diseases as well as premature death [911]. Incon-
sistent conclusions across studies might amongst other reasons arrive from unreliable measure-
ments of SED, as most of these studies have included 35 days of measurement [1217], with
some exceptions (1 day [18]; 67 days [19, 20]). Moreover, as SED are likely to be related
to wear time, correction for wear time might improve reliability. Consistent with this hypothe-
sis, percent SED has previously been shown to be superior to minutes of SED as a predictor of
metabolic risk [18].
Historically, sedentary behavior (SED) was conceptualized as the lower end of the PA
spectrum, as opposed to moderate- to vigorous PA (MVPA), but is now increasingly being
viewed as a behavior distinct from PA, defined as waking behavior characterized by an energy
expenditure 1.5 metabolic equivalents (METs) [21]. Although somewhat arbitrary, studies
using Actigraph instruments mainly operationalize SED as the time spent <100 cpm. This cut-
off differs substantially from that used to define inactivity (<500 cpm) applied by Matthews
et al [6]. Thus, besides the study of reliability of SED in older adults [8], to the best of our
knowledge, reliability of SED obtained by accelerometry have not been investigated in adults,
and should be prioritized [22]. Furthermore, no studies have determined the intra-individual
week-by-week agreement of accelerometer outcomes using absolute measures of reliability (i.e.,
standard error of the measurement and limits of agreement). Most evidence suggest that a reli-
ability of 0.700.80 are achieved for PA with 37 days of monitoring by estimation of the
number of days needed based on the Spearman Brown prophecy formula, when measurements
are conducted over a single 7-day period [57, 2, 8]. However, such study designs have received
critique for possibly leaving to optimistic results and should be interpreted with caution [23
25]. First, the results are in principle only generalizable to the included days, as inclusion of
additional days, weeks or seasons will add variability. Secondly, the assumption of compound
symmetry (i.e., similar variances and co-variances) across days of measurement might not be
fulfilled. Additionally, ICC is the variance partitioning of subjects to the total variance, thus
ICC is a relative and context-specific estimate that depends on the heterogeneity of the sample
[2628]. Thus, research targeting agreement of SED and PA measurements by means of abso-
lute measures of reliability, which allow for a direct quantification of how much outcomes vary
over time independent of the variability of observations should be given priority.
The aim of the present study was to determine the agreement of objectively assessed SED
and PA over 3 subsequent weeks in an adult population. Based on previous findings [8, 6], we
hypothesized that a longer monitoring period would be needed to obtain reliable estimates of
SED, compared to other outcomes.
Procedures
Physical activity was measured using the Actigraph GT3X+ accelerometer (firmware 2.2.1)
(Pensacola, FL, USA). Two accelerometers were attached to an elastic belt, and worn at contra-
lateral hips over a period of 21 days. Half-way through the measurement period (day 11), sub-
jects switched the accelerometers units around (1 unit was worn on the right side the first 10
days and at the left side the last 10 days, and vice versa), two allow for a within-accelerometer
controlled study off contralateral hip differences. As differences between the hips were abso-
lutely minimal (effect size 0.04), all results are based on output from one of the units for the
purpose of the present study. Subjects were instructed to wear the accelerometer at all times,
except during water activities (swimming, showering) or while sleeping. Units were initialized
at a sampling rate of 30 Hz. Files were analyzed at 10 second epochs using the Kinesoft v.
3.3.75 software [29]. Data was analyzed using different criteria for valid wear time (8; 10;
12 hours/day). In all analyses, consecutive periods of 60 minutes of zero counts (allowing
for 2 minutes of non-zero counts) were defined as non-wear time [2, 30] and excluded prior
to scoring.
Results were initially reported for overall PA level (CPM), as well as SED (<100 cpm), light
PA (LPA) (1002019 cpm), MPA (20205998 cpm), VPA (5999 cpm) and MVPA (2020
cpm) obtained from the vertical axis (axis 1) [31], as well as the vector magnitude (VM) CPM.
However, as the level of VPA were very low (median (interquartile range) 1.3 (9.0) min/day),
MPA and VPA were collapsed and only MVPA were reported throughout the study. Intensity-
specific PA and SED were reported as minutes/day and as percentage values of daily valid wear
time.
Subjects characteristics (sex, age, body mass and height) were self-reported. Body mass
index (BMI) was calculated as the body mass divided by the squared height (kg/m2).
Statistical analyses
Subject characteristics were reported as means and standard deviations (SD).
The association between wear time and PA and SED were investigated using linear mixed
model regression analyses including a random intercept for subjects. Differences between
week- and weekend days were tested by including a dummy variable (week- vs. weekend day)
in the above-mentioned model (i.e., adjustment were made for wear time). Residuals were nor-
mally distributed in all models. Findings were reported as regression coefficients and 95% con-
fidence intervals (CI). Further, we used variance partitioning obtained from a two-way mixed
effect model reporting the ratio of week-weekend variance to the total variance (week-weekend
variance/(between subject variance + week-weekend variance + residual variance)), including
wear time as a fixed effect [32]. A wear time criterion of 10 hours/day was used for these
analyses.
The single-day reliability and number of days needed to obtain the desired reliability was
assessed for wear times of 8; 10; 12 hours/day. Reliability for single days of measurement
was assessed using variance partitioning obtained through a one-way random effect model,
reporting the ratio of the between subject variance to the total variance (between subject
variance/(between subject variance + residual variance)) [26]. Number of days needed to
obtain a reliability of 0.80 was estimated using the Spearman Brown prophecy formula [2, 32]:
N = ICCt/(1-ICCt) [(1-ICCs)/ICCs], where N = number of days needed, ICCt = desired level of
reliability, and ICCs = reliability for single days.
Bland Altman plots, showing the difference between two subsequent weeks as a function
of the mean of the two weeks [26], was applied to determine the week-by-week variability.
Because the data generally were deemed to be homoscedastic (y not related to x), the standard
error of the measurement (SEM) and limits of agreement (LoA) were calculated from the resid-
ual variance (i.e., within-subjects) error term obtained through a one-way random effect model
using the 3 accumulated weekly means (as opposed to 21 single day observations) as the depen-
p p
dent variable (SEM = residual variance; LoA = SEM 2 1.96) [28]. As the pattern of vari-
ability for overall PA and MVPA were heteroscedastic (y related to x), we also reported the
percentage typical error (SEM/median values 100), which are more appropriate when variabil-
ity depends on the value of the measurement. Finally, results were also reported with adjust-
ment for wear time by including wear time as a fixed effect in a two-way mixed effect model
[32].
All analyses were performed using IBM SPSS v. 20 (IBM SPSS Statistics for Windows,
Armonk, NY: IBM Corp., USA). A p-value < .05 indicated statistically significant findings.
Results
Subject characteristics
All 87 subjects (28 (32%) men and 59 women; age 31.3 (12.2) years; body mass 70.4 (12.2) kg,
height 172.1 (8.1) cm, body mass index 23.7 (3.1) kg/m2) was included in the analyses. Weekly
mean (SD) SED and PA across the three weeks were: CPM = 486 (172); VM CPM = 878 (242);
SED = 564 (63) min/day / 69.5 (6.2)%; LPA = 183 (57) min/day / 22.3 (5.8)%; MVPA = 65.9
(28.1) min/day / 8.2 (3.6)%.
The association between wear time and the different outcome variables for physical activity and sedentary
behavior.
CPM = counts per minute; VM CPM = vector magnitude counts per minute; SED = sedentary time;
LPA = light physical activity; MVPA = moderate-to-vigorous physical activity; CI = condence interval
doi:10.1371/journal.pone.0133296.t001
wear time was negatively related to CPM and percentage time in MVPA, whereas no relation-
ship was found for VM CPM.
Reliability for single days of measurement (ICCs) and number of days needed to achieve a reliability of 0.80 (N) for different criteria to determine a valid
day of measurement.
CPM = counts per minute; VM CPM = vector magnitude counts per minute; SED = sedentary time; LPA = light physical activity; MVPA = moderate-to-
vigorous physical activity; ICCs = ICC for a single day of measurement; N = number of days needed to achieve a ICC = 0.80
doi:10.1371/journal.pone.0133296.t002
substantial variation in wear time across weeks (SEM (%) / LoA = 46.6 (5.7) / 129.1 min/day),
given the relationships with LPA and SED as shown previously (Table 1). Other variables were
more or less unaffected by the adjustment for wear time. Although nearly all variables reached
The week-by-week reliability for different outcome variables for three consecutive weeks of measurement without- and with adjustment for wear time,
using wear time criteria of 10 hours/day and 4 valid days.
CPM = counts per minute; VM CPM = vector magnitude counts per minute; SED = sedentary time; LPA = light physical activity; MVPA = moderate-to-
vigorous physical activity; ICCs = ICC for a single day of measurement; SEM (%) = standard error of the measurement (percentage of the mean value);
LoA = limits of agreement
1
Analyzed by means of a one-way random effect model
2
Analyzed by means of a two-way mixed effect model.
doi:10.1371/journal.pone.0133296.t003
Fig 1. Bland Altman plots of measurement variability for different outcome variables. Bland Altman plots (the difference between two weeks of
measurement on the y-axis versus the mean of the two weeks on the x-axis) for wear time (a); average counts per minute (CPM) from the vertical axis (b) and
the vector magnitude (VM CPM) (c); absolute time spent sedentary (SED) (d), light physical activity (LPA) (e) and moderate-to-vigorous physical activity
(MVPA) (f); percentage time in SED (g), LPA (h) and MVPA (i). Results are based on a 10 hours & 4 days wear time criteria. Filled dots are week 1week
2 (n = 59); open squares are week 1week 3 (n = 53)95% limits of agreement are indicated as dotted lines.
doi:10.1371/journal.pone.0133296.g001
an ICC 0.70, the absolute measures of reliability clearly shows that a substantial degree of
individual variability must be expected across subsequent weeks. After adjustment for wear
time, LoAs was 1.01.6 times the sample SDs. When typical errors were reported as a
The week-by-week reliability for different outcome variables for different criteria to determine a valid day- and week of measurement.
CPM = counts per minute; VM CPM = vector magnitude counts per minute; SED = sedentary time; LPA = light physical activity; MVPA = moderate-to-
vigorous physical activity; ICCs = ICC for a single day of measurement; SEM = standard error of the measurement; Limits of agreement can be calculated
p
as SEM* 2*1.96
doi:10.1371/journal.pone.0133296.t004
percentage value of the mean, the greatest variability vas found for MVPA (~24%) and the low-
est for SED (~5%).
We found slight improvements in week-by-week reliability when data was accumulated
over longer (8 to 12 h) and more days (3 to 5 d), however, the pattern was not very consistent
and the differences were minor (Table 4).
Discussion
The present study is the first to investigate the week-by-week reliability of measurements of
SED and PA, as obtained by accelerometry in adults, using absolute measures of reliability.
There was a clear difference in reliability for SED reported in terms of minutes/day (absolute
measure) compared to as a percentage value of the valid wear time (relative measure), favoring
the latter. This disparate finding is caused by the strong relationship between wear time and
SED. Thus, adjustment for wear time reduced the variability in SED (min/day) considerably
and left similar ICC as for SED (%). Absolute measures of reliability showed that week-by-
week differences must be expected to be substantial for all outcome variables. Contrary to our
hypothesis, the reliability of SED and PA were similar, or, when variability was reported rela-
tive to the mean values; considerably better for SED compared to PA.
Although wear time criteria differ between studies, many studies employ a wear time criteria
of 10 hours/day for 35 days/week to constitute a valid accelerometer measurement in
adults [22, 16, 14, 17, 13, 12]. These criteria are in line with current recommendations [22, 2].
The present findings suggest that the number of days to be considered a valid measurement
might be increased to a full week or longer to achieve an ICC of 0.80, which is currently the
most applied desired reliability level. Consistent with previous studies [6, 8], we found
impaired reliability for SED (min/day) compared to PA. This could indicate that SED is a
behavior that are difficult to capture as it might vary more on a day-by-day basis than other
activities that are performed on a higher intensity-level. However, SED given as a relative mea-
sure (percentage of valid wear time) was superior to SED reported in absolute term (minutes/
day) (requiring 7 vs. 15 days of measurement for the 10 h criteria) in the present study. This
finding results from a considerable day-to-day variation in wear time, given the very strong
relationship between wear time and SED (min/day) shown in the present study. Using a 10 h
4d criteria, wear time must be expected to differ by >2 hours/day between subsequent weeks.
Given the strong relationship between wear time and SED ( = 0.709: i.e., 1 minute change in
wear time would translate into ~43 sec change in SED time), adjustment for wear time reduced
the variability for SED by ~37%. A relatively strong relationship with wear time was also
shown for LPA, whereas weaker (although statistically significant) relationships were found for
other variables. This clearly indicates that wear time should be considered a crucial variable in
data reduction and analysis of accelerometer data, which is consistent with findings of a recent
study [33]. Herrmann et al [33] showed patterns across 10 to 14 hours of wear time in a simula-
tion study based on NHANES data which coincide with the present findings. Thus, our find-
ings, along with those of others, shows that SED and PA as determined by accelerometry,
should be either reported as percentage values or, alternatively, analyses could be adjusted for
wear time, which might be preferred if appropriate. Studies that have investigated the associa-
tion between SED and cardio-metabolic health vary considerable in their applied operationali-
zation of SED, using absolute values without adjustment for wear time [14, 19], absolute values
with adjustment for wear time [15, 17, 20], percentage values [16, 13] or a combination of sev-
eral approaches [18, 12]. Thus, variation in reliability might be one source to explain disparate
findings across studies.
The present estimates of number of days needed to achieve a reliability of ICC = 0.80 for
overall PA and MVPA, fall in between of previous estimates. While several studies have found
that 26 days are required [5, 7, 8, 6, 30], other studies have shown that 12 [4] and 1623 days
are needed [3]. Jerome et al [3] used a wear time criteria of 6 h, which might have led to the
need for many days of measurement, in line with the trend across the wear time criteria applied
in the present study. Levin et al [4] applied a protocol with measurements during an entire
year, which may have introduced greater variation (e.g., seasonal effects) than across subse-
quent days and weeks. Nevertheless, as ICC is based on variance partitioning and depends on
the relative difference in variances (inter- vs. intra-individual sources of variance), the coeffi-
cient is context-specific. Thus, it change with the range of observations (similar to the correla-
tion coefficient), despite residual variance and absolute measures of reliability being unaffected
[2628]. The current sample was a convenience sample exhibiting a higher activity level (but
with approximately similar heterogeneity as evaluated by SDs for the different outcome vari-
ables) as compared to population estimates for the corresponding age group [3436]. Thus, it
could be argued that our estimates of ICC might be fairly generalizable to the general adult
population. Yet, previous studies have concluded that it does not seem to be any relationship
between activity level and reliability [6, 4]. Based on the heteroscedastic patterns of variability
found for overall PA level and MVPA (Fig 1), the current study indicates otherwise. Thus, the
sample characteristics (exhibiting a high PA level) might be a likely explanation for the lower
reliability found for PA compared to other studies. This means that researchers should deter-
mine reliability in their specific samples to make appropriate decisions on valid wear criteria in
a given study. Yet, it could be hypothesized from the current study that SED are less influenced
by sample characteristics than PA, as a homoscedastic pattern was evident for SED as opposed
to the heteroscedastic pattern found for PA.
As expected, the findings presented in Table 2 suggest that increased wear time (both hours
per day and the total number of days) will lead to improved reliability. However, this pattern
was less pronounced when different wear time criteria was applied to the week-by-week analy-
ses. This finding is consistent with the finding by Jerome et al [3], who detected equivalent
estimates of reliability for wear times of 6 and 10 h/day, as well as for 4 and 7 days of moni-
toring. Based on this finding, Jerome et al concluded that Spearman-Brown estimates should
be interpreted with caution and be secondary to actual reliability estimates. We have no other
explanation for such disparate findings than suggesting that averaging several days of measure-
ment can lead to relatively stable estimates of weekly SED and PA despite great variability
between days, thus, weekly activity might not be properly modeled by the Spearman-Brown
formula. Together, these findings suggest that future studies should determine the week-by-
week test-retest reliability of accelerometer measurements, and to a lesser extent rely on esti-
mates based on day-by-day measurements.
Nevertheless, our findings suggest that 69 days of measurement might be needed to reli-
ably capture SED and PA, given a wear time criteria of 1012 h/day. This necessitates an
increased monitoring length, compared to that of 7 days that is usually applied. Although this
might improve validity of study conclusions, the burden for subjects should be kept minimal to
maximize response rate and adherence to the monitoring protocol. Feedback from our subjects
clearly indicated that 21 days of monitoring was a lengthy measurement. However, 14 days
would be appropriate, under the assumption that subjects obtain 45 valid days per week. Still,
this is a matter of the research question posed, as population-estimates on a group level
requires less reliability than individual-level estimates. Also, an increased wear time per day is
an option. However, the number of observations left for analyses declined with increased wear
time criteria, as shown previously [37]. Thus, the choice of wear time criteria is a trade-off
between reliability and power, of which both are of crucial importance to arrive at valid conclu-
sions, as the loss of both will bias conclusions towards the null-hypothesis. As noise in predic-
tor (x-) variables will lead to attenuation of regression coefficients by biasing coefficients
towards the null (regression dilution bias), and noise in outcome (y-) variables will increase
standard errors and increase the likelihood of performing type II errors [1], unreliable mea-
sures will weaken researchers ability to make valid conclusions in PA epidemiology. If PA is
modeled as an exposure variable, increased sample size on the expense of reliability will only
make the results more precisely wrong [1].
We found significant, although small, differences in activity level and SED between week
days and weekend days. Although previous studies have shown somewhat conflicting findings
on this point [7, 8, 6], application of a wear day criteria requiring the inclusion of at least one
weekend day may be justifiable based on the current study.
Conclusions
We conclude that more than 7 days of measurement was needed to achieve a reliability (ICC)
of 0.80 in the current study, depending on the variable(s) of interest. Although we obtained a
reliability of ICC >0.70 for 3 subsequent weeks of measurement using frequently applied crite-
ria for wear time, considerable week-by-week variability was found. The findings clearly dem-
onstrated that wear time was of crucial importance to reliably assess SED. Thus, contrary to
our hypothesis, the reliability of assessing SED and PA were similar, when wear time was
appropriately corrected. Researchers need to be aware of intra-individual variability in acceler-
ometer-measurements and take appropriate actions according to the hypothesis under study,
because noise in any measurement will attenuate real relationships between PA and health
and increase the likelihood of performing type II errors. In terms of the number of monitoring
days needed, this is a trade-off between reliability and statistical power.
Supporting Information
S1 Dataset. Supplementary data file including all material underlying the present study.
(SAV)
Acknowledgments
We thank all participants that made the study possible.
Author Contributions
Conceived and designed the experiments: EA EY. Performed the experiments: EY. Analyzed
the data: EA. Wrote the paper: EA EY.
References
1. Hutcheon JA, Chiolero A, Hanley JA. Random measurement error and regression dilution bias. BMJ.
2010; 340: 140206
2. Trost SG, McIver KL, Pate RR. Conducting accelerometer-based activity assessments in field-based
research. Med Sci Sports Exerc. 2005; 37(11):S531S43. PMID: 16294116
3. Jerome GJ, Young DR, Laferriere D, Chen CH, Vollmer WM. Reliability of RT3 accelerometers among
overweight and obese adults. Med Sci Sports Exerc. 2009; 41(1):1104. doi: 10.1249/MSS.
0b013e3181846cd8 PMID: 19092700
4. Levin S, Jacobs DR, Ainsworth BE, Richardson MT, Leon AS. Intra-individual variation and estimates
of usual physical activity. Ann Epidemiol. 1999; 9(8):4818. PMID: 10549881
5. Coleman KJ, Epstein LH. Application of generalizability theory to measurement of activity in males who
are not regularly active: A preliminary report. Res Q Exerc Sport. 1998; 69(1):5863. PMID: 9532623
6. Matthews CE, Ainsworth BE, Thompson RW, Bassett DR. Sources of variance in daily physical activity
levels as measured by an accelerometer. Med Sci Sports Exerc. 2002; 34(8):137681. PMID:
12165695
7. Gretebeck RJ, Montoye HJ. Variability of some objective measures of physical activity. Med Sci Sports
Exerc. 1992; 24(10):116772. PMID: 1435166
8. Hart TL, Swartz AM, Cashin SE, Strath SJ. How many days of monitoring predict physical activity and
sedentary behaviour in older adults? Int J Behav Nutr Phys Act. 2011; 8: Article ID 62.
9. Ford ES, Caspersen CJ. Sedentary behaviour and cardiovascular disease: a review of prospective
studies. Int J Epidemiol. 2012; 41(5):133853. doi: 10.1093/ije/dys078 PMID: 22634869
10. Edwardson CL, Gorely T, Davies MJ, Gray LJ, Khunti K, Wilmot EG et al. Association of sedentary
behaviour with metabolic syndrome: A Meta-Analysis. Plos One. 2012; 7(4).
11. Wilmot EG, Edwardson CL, Achana FA, Davies MJ, Gorely T, Gray LJ et al. Sedentary time in adults
and the association with diabetes, cardiovascular disease and death: systematic review and meta-anal-
ysis. Diabetologia. 2012; 55(11):2895905. doi: 10.1007/s00125-012-2677-z PMID: 22890825
12. McGuire KA, Ross R. Sedentary behavior is not associated with cardiometabolic risk in adults with
abdominal obesity. Plos One. 2011; 6(6).
13. Healy GN, Wijndaele K, Dunstan DW, Shaw JE, Salmon J, Zimmet PZ et al. Objectively measured sed-
entary time, physical activity, and metabolic risk: The Australian Diabetes, Obesity and Lifestyle Study
(AusDiab). Diabetes Care. 2008; 31(2):36971. PMID: 18000181
14. Celis-Morales CA, Perez-Bravo F, Ibanez L, Salas C, Bailey MES, Gill JMR. Objective vs. self-reported
physical activity and sedentary time: Effects of measurement method on relationships with risk bio-
markers. Plos One. 2012; 7(5).
15. Ekelund U, Griffin SJ, Wareham NJ, ProActive Res G. Physical activity and metabolic risk in individuals
with a family history of type 2 diabetes. Diabetes Care. 2007; 30(2):33742. PMID: 17259504
16. Aadland E, Andersen JR, Anderssen SA, Kvalheim OM. Physical activity versus sedentary behavior:
Associations with lipoprotein particle subclass concentrations in healthy adults. Plos One. 2013; 8(12).
17. Green AN, McGrath R, Martinez V, Taylor K, Paul DR, Vella CA. Associations of objectively measured
sedentary behavior, light activity, and markers of cardiometabolic health in young women. Eur J Appl
Physiol. 2014; 114(5):90719. doi: 10.1007/s00421-014-2822-0 PMID: 24463602
18. Bankoski A, Harris TB, McClain JJ, Brychta RJ, Caserotti P, Chen KY et al. Sedentary activity associ-
ated with metabolic syndrome independent of physical activity. Diabetes Care. 2011; 34(2):497503.
doi: 10.2337/dc10-0987 PMID: 21270206
19. Scheers T, Philippaerts R, Lefevre J. SenseWear-determined physical activity and sedentary behavior
and metabolic syndrome. Med Sci Sports Exerc. 2013; 45(3):4819. doi: 10.1249/MSS.
0b013e31827563ba PMID: 23034646
20. Kim J, Tanabe K, Yokoyama N, Zempo H, Kuno S. Objectively measured light-intensity lifestyle activity
and sedentary time are independently associated with metabolic syndrome: a cross-sectional study of
Japanese adults. Int J Behav Nutr Phys Act. 2013; 10: Article ID 30.
21. Barnes J, Behrens TK, Benden ME, Biddle S, Bond D, Brassard P et al. Letter to the Editor: Standard-
ized use of the terms "sedentary" and "sedentary behaviours". Appl Physiol Nutr Metab. 2012; 37
(3):5402. doi: 10.1139/h2012-024 PMID: 22540258
22. Atkin AJ, Gorely T, Clemes SA, Yates T, Edwardson C, Brage S et al. Methods of measurement in epi-
demiology: Sedentary behaviour. Int J Epidemiol. 2012; 41(5):146071. doi: 10.1093/ije/dys118 PMID:
23045206
23. Baranowski T, Masse LC, Ragan B, Welk G. How many days was that? We're still not sure, but we're
asking the question better! Med Sci Sports Exerc. 2008; 40(7):S544S9. doi: 10.1249/MSS.
0b013e31817c6651 PMID: 18562972
24. Matthews CE, Hagstromer M, Pober DM, Bowles HR. Best practices for using physical activity monitors
in population-based research. Med Sci Sports Exerc. 2012; 44:S68S76. doi: 10.1249/MSS.
0b013e3182399e5b PMID: 22157777
25. Wickel EE, Welk GJ. Applying generalizability theory to estimate habitual activity levels. Med Sci Sports
Exerc. 2010; 42(8):152834. doi: 10.1249/MSS.0b013e3181d107c4 PMID: 20139788
26. Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical
measurement. Lancet. 1986; 1(8476):30710. PMID: 2868172
27. Hopkins WG. Measures of reliability in sports medicine and science. Sports Med. 2000; 30(1):115.
PMID: 10907753
28. Weir JP. Quantifying test-retest reliability using the intraclass correlation coefficient and the SEM. J
Strength Cond Res. 2005; 19(1):23140. PMID: 15705040
29. Kinesoft Software . Available from http://kinesoft.org/ (accessed 28th of November 2014).
30. Sirard JR, Forsyth A, Oakes JM, Schmitz KH. Accelerometer test-retest reliability by data processing
algorithms: Results from the Twin Cities Walking Study. J Phys Act Health. 2011; 8(5):66874. PMID:
21734312
31. Metzger JS, Catellier DJ, Evenson KR, Treuth MS, Rosamond WD, Siega-Riz AM. Patterns of objec-
tively measured physical activity in the United States. Med Sci Sports Exerc. 2008; 40(4):6308. doi:
10.1249/MSS.0b013e3181620ebc PMID: 18317384
32. McGraw KO, Wong SP. Forming inferences about some intraclass correlation coefficients. Psychol
Methods. 1996; 1(1):3046.
33. Herrmann SD, Barreira TV, Kang M, Ainsworth BE. Impact of accelerometer wear time on physical
activity data: a NHANES semisimulation data approach. Br J Sports Med. 2014; 48(3):27882. doi: 10.
1136/bjsports-2012-091410 PMID: 22936409
34. Hansen BH, Kolle E, Dyrstad SM, Holme I, Anderssen SA. Accelerometer-determined physical activity
in adults and older people. Med Sci Sports Exerc. 2012; 44(2):26672. doi: 10.1249/MSS.
0b013e31822cb354 PMID: 21796052
35. Troiano RP, Berrigan D, Dodd KW, Masse LC, Tilert T, McDowell M. Physical activity in the United
States measured by accelerometer. Med Sci Sports Exerc. 2008; 40(1):1818. PMID: 18091006
36. Hagstrmer M, Oja P, Sjstrm M. Physical activity and inactivity in an adult population assessed by
accelerometry. Med Sci Sports Exerc. 2007; 39(9):15028. PMID: 17805081
37. Colley R, Connor Gorber S, Tremblay MS. Quality control and data reduction procedures for accelero-
metry-derived measures of physical activity. Health Reports. 2010; 21(1):639. PMID: 20426228